Skip to content

Commit 69dd6bc

Browse files
k8s-ci-robotk8s-ci-robot
authored
Merge pull request #51606 from windsonsea/setuper
[zh] Add set-up-dra-cluster.md
2 parents 9db67fa + 8c3579e commit 69dd6bc

File tree

2 files changed

+316
-0
lines changed

2 files changed

+316
-0
lines changed

‎content/zh-cn/docs/tasks/configure-pod-container/assign-resources/set-up-dra-cluster.md‎

Lines changed: 307 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,307 @@
1+
---
2+
title: "在集群中设置 DRA"
3+
content_type: task
4+
min-kubernetes-server-version: v1.32
5+
weight: 10
6+
---
7+
<!--
8+
title: "Set Up DRA in a Cluster"
9+
content_type: task
10+
min-kubernetes-server-version: v1.32
11+
weight: 10
12+
-->
13+
14+
{{< feature-state feature_gate_name="DynamicResourceAllocation" >}}
15+
16+
<!-- overview -->
17+
18+
<!--
19+
This page shows you how to configure _dynamic resource allocation (DRA)_ in a
20+
Kubernetes cluster by enabling API groups and configuring classes of devices.
21+
These instructions are for cluster administrators.
22+
-->
23+
本文介绍如何在 Kubernetes 集群中通过启用 API 组并配置设备类别来设置**动态资源分配（DRA）**。
24+
这些指示说明适用于集群管理员。
25+
26+
<!-- body -->
27+
28+
<!--
29+
## About DRA {#about-dra}
30+
-->
31+
## 关于 DRA {#about-dra}
32+
33+
{{< glossary_definition term_id="dra" length="all" >}}
34+
35+
<!--
36+
Ensure that you're familiar with how DRA works and with DRA terminology like
37+
{{< glossary_tooltip text="DeviceClasses" term_id="deviceclass" >}},
38+
{{< glossary_tooltip text="ResourceClaims" term_id="resourceclaim" >}}, and
39+
{{< glossary_tooltip text="ResourceClaimTemplates" term_id="resourceclaimtemplate" >}}.
40+
For details, see
41+
[Dynamic Resource Allocation (DRA)](/docs/concepts/scheduling-eviction/dynamic-resource-allocation/).
42+
-->
43+
确保你已了解 DRA 的工作机制及其术语，例如
44+
{{< glossary_tooltip text="DeviceClasses" term_id="deviceclass" >}}、
45+
{{< glossary_tooltip text="ResourceClaims" term_id="resourceclaim" >}}以及
46+
{{< glossary_tooltip text="ResourceClaimTemplates" term_id="resourceclaimtemplate" >}}。
47+
更多信息请参见[动态资源分配（DRA）](/zh-cn/docs/concepts/scheduling-eviction/dynamic-resource-allocation/)。
48+
49+
<!-- prerequisites -->
50+
51+
## {{% heading "prerequisites" %}}
52+
53+
{{< include "task-tutorial-prereqs.md" >}} {{< version-check >}}
54+
55+
<!--
56+
* Directly or indirectly attach devices to your cluster. To avoid potential
57+
issues with drivers, wait until you set up the DRA feature for your
58+
cluster before you install drivers.
59+
-->
60+
* 将设备直接或间接挂接到你的集群中。为避免驱动相关的问题，请在安装驱动之前先完成 DRA 特性的配置。
61+
62+
<!-- steps -->
63+
64+
<!--
65+
## Enable the DRA API groups {#enable-dra}
66+
67+
To let Kubernetes allocate resources to your Pods with DRA, complete the
68+
following configuration steps:
69+
70+
1. Enable the `DynamicResourceAllocation`
71+
[feature gate](/docs/reference/command-line-tools-reference/feature-gates/)
72+
on all of the following components:
73+
-->
74+
## 启用 DRA API 组 {#enable-dra}
75+
76+
若要让 Kubernetes 能够使用 DRA 为你的 Pod 分配资源，需完成以下配置步骤：
77+
78+
1. 在所有以下组件中启用 `DynamicResourceAllocation`
79+
[特性门控](/zh-cn/docs/reference/command-line-tools-reference/feature-gates/)：
80+
81+
* `kube-apiserver`
82+
* `kube-controller-manager`
83+
* `kube-scheduler`
84+
* `kubelet`
85+
86+
<!--
87+
1. Enable the following
88+
{{< glossary_tooltip text="API groups" term_id="api-group" >}}:
89+
90+
* `resource.k8s.io/v1beta1`: required for DRA to function.
91+
* `resource.k8s.io/v1beta2`: optional, recommended improvements to the user
92+
experience.
93+
94+
For more information, see
95+
[Enabling or disabling API groups](/docs/reference/using-api/#enabling-or-disabling).
96+
-->
97+
2. 启用以下 {{< glossary_tooltip text="API 组" term_id="api-group" >}}：
98+
99+
* `resource.k8s.io/v1beta1`：DRA 所必需。
100+
* `resource.k8s.io/v1beta2`：可选，推荐启用以提升用户体验。
101+
102+
更多信息请参阅[启用或禁用 API 组](/zh-cn/docs/reference/using-api/#enabling-or-disabling)。
103+
104+
<!--
105+
## Verify that DRA is enabled {#verify}
106+
107+
To verify that the cluster is configured correctly, try to list DeviceClasses:
108+
-->
109+
## 验证是否启用了 DRA {#verify}
110+
111+
若要验证集群是否配置正确，可尝试列出 DeviceClass：
112+
113+
```shell
114+
kubectl get deviceclasses
115+
```
116+
117+
<!--
118+
If the component configuration was correct, the output is similar to the
119+
following:
120+
-->
121+
如果组件配置正确，输出类似如下：
122+
123+
```
124+
No resources found
125+
```
126+
127+
<!--
128+
If DRA isn't correctly configured, the output of the preceding command is
129+
similar to the following:
130+
-->
131+
如果 DRA 未正确配置，则上述命令的输出可能如下：
132+
133+
```
134+
error: the server doesn't have a resource type "deviceclasses"
135+
```
136+
137+
<!--
138+
Try the following troubleshooting steps:
139+
140+
1. Ensure that the `kube-scheduler` component has the `DynamicResourceAllocation`
141+
feature gate enabled *and* uses the
142+
[v1 configuration API](/docs/reference/config-api/kube-scheduler-config.v1/).
143+
If you use a custom configuration, you might need to perform additional steps
144+
to enable the `DynamicResource` plugin.
145+
1. Restart the `kube-apiserver` component and the `kube-controller-manager`
146+
component to propagate the API group changes.
147+
-->
148+
你可以尝试以下排查步骤：
149+
150+
1. 确保 `kube-scheduler` 组件已启用 `DynamicResourceAllocation` 特性门控，并且使用的是
151+
[v1 配置 API](/zh-cn/docs/reference/config-api/kube-scheduler-config.v1/)。
152+
如果你使用自定义配置，你可能还需额外启用 `DynamicResource` 插件。
153+
154+
2. 重启 `kube-apiserver` 和 `kube-controller-manager` 组件，以传播 API 组变更。
155+
156+
<!--
157+
## Install device drivers {#install-drivers}
158+
159+
After you enable DRA for your cluster, you can install the drivers for your
160+
attached devices. For instructions, check the documentation of your device
161+
owner or the project that maintains the device drivers. The drivers that you
162+
install must be compatible with DRA.
163+
164+
To verify that your installed drivers are working as expected, list
165+
ResourceSlices in your cluster:
166+
-->
167+
## 安装设备驱动 {#install-drivers}
168+
169+
你启用集群的 DRA 特性后，你可以安装所挂接设备的驱动。
170+
安装方式请参见设备所有者或驱动维护方提供的文档。你安装的驱动必须与 DRA 兼容。
171+
172+
若要验证驱动是否正常工作，可列出集群中的 ResourceSlice：
173+
174+
```shell
175+
kubectl get resourceslices
176+
```
177+
178+
<!--
179+
The output is similar to the following:
180+
-->
181+
输出示例如下：
182+
183+
```
184+
NAME NODE DRIVER POOL AGE
185+
cluster-1-device-pool-1-driver.example.com-lqx8x cluster-1-node-1 driver.example.com cluster-1-device-pool-1-r1gc 7s
186+
cluster-1-device-pool-2-driver.example.com-29t7b cluster-1-node-2 driver.example.com cluster-1-device-pool-2-446z 8s
187+
```
188+
189+
<!--
190+
## Create DeviceClasses {#create-deviceclasses}
191+
192+
You can define categories of devices that your application operators can
193+
claim in workloads by creating
194+
{{< glossary_tooltip text="DeviceClasses" term_id="deviceclass" >}}. Some device
195+
driver providers might also instruct you to create DeviceClasses during driver
196+
installation.
197+
-->
198+
## 创建 DeviceClass {#create-deviceclasses}
199+
200+
你可以通过创建
201+
{{< glossary_tooltip text="DeviceClasses" term_id="deviceclass" >}}
202+
定义设备的分类，供应用运维人员在工作负载中申领这些设备。
203+
某些设备驱动提供方也可能在驱动安装过程中要求你创建 DeviceClass。
204+
205+
<!--
206+
The ResourceSlices that your driver publishes contain information about the
207+
devices that the driver manages, such as capacity, metadata, and attributes. You
208+
can use {{< glossary_tooltip term_id="cel" >}} to filter for properties in your
209+
DeviceClasses, which can make finding devices easier for your workload
210+
operators.
211+
212+
1. To find the device properties that you can select in DeviceClasses by using
213+
CEL expressions, get the specification of a ResourceSlice:
214+
-->
215+
你的驱动所发布的 ResourceSlice 中包含了设备的相关信息，例如容量、元数据和属性。你可以使用
216+
{{< glossary_tooltip term_id="cel" >}} 表达式按 DeviceClass 中的属性进行筛选，
217+
从而帮助工作负载运维人员更轻松地找到合适的设备。
218+
219+
1. 若要查看可通过 CEL 表达式在 DeviceClass 中选择的设备属性，你可以查看某个 ResourceSlice 的规约：
220+
221+
```shell
222+
kubectl get resourceslice <resourceslice-name> -o yaml
223+
```
224+
225+
<!--
226+
The output is similar to the following:
227+
-->
228+
229+
输出类似如下：
230+
231+
<!--
232+
# lines omitted for clarity
233+
-->
234+
235+
```yaml
236+
apiVersion: resource.k8s.io/v1beta1
237+
kind: ResourceSlice
238+
# 为简洁省略部分内容
239+
spec:
240+
devices:
241+
- basic:
242+
attributes:
243+
type:
244+
string: gpu
245+
capacity:
246+
memory:
247+
value: 64Gi
248+
name: gpu-0
249+
- basic:
250+
attributes:
251+
type:
252+
string: gpu
253+
capacity:
254+
memory:
255+
value: 64Gi
256+
name: gpu-1
257+
driver: driver.example.com
258+
nodeName: cluster-1-node-1
259+
# 为简洁省略部分内容
260+
```
261+
262+
<!--
263+
You can also check the driver provider's documentation for available
264+
properties and values.
265+
-->
266+
267+
你也可以查阅驱动提供商的文档，了解可用的属性和对应值。
268+
269+
<!--
270+
1. Review the following example DeviceClass manifest, which selects any device
271+
that's managed by the `driver.example.com` device driver:
272+
-->
273+
2. 查看以下 DeviceClass 示例清单，它选择所有由 `driver.example.com` 设备驱动管理的设备：
274+
275+
{{% code_sample file="dra/deviceclass.yaml" %}}
276+
277+
<!--
278+
1. Create the DeviceClass in your cluster:
279+
-->
280+
3. 在集群中创建 DeviceClass：
281+
282+
```shell
283+
kubectl apply -f https://k8s.io/examples/dra/deviceclass.yaml
284+
```
285+
286+
<!--
287+
## Clean up {#clean-up}
288+
289+
To delete the DeviceClass that you created in this task, run the following
290+
command:
291+
-->
292+
## 清理 {#clean-up}
293+
294+
要删除本任务中创建的 DeviceClass，运行以下命令：
295+
296+
```shell
297+
kubectl delete -f https://k8s.io/examples/dra/deviceclass.yaml
298+
```
299+
300+
## {{% heading "whatsnext" %}}
301+
302+
<!--
303+
* [Learn more about DRA](/docs/concepts/scheduling-eviction/dynamic-resource-allocation)
304+
* [Allocate Devices to Workloads with DRA](/docs/tasks/configure-pod-container/assign-resources/allocate-devices-dra)
305+
-->
306+
* [进一步了解 DRA](/zh-cn/docs/concepts/scheduling-eviction/dynamic-resource-allocation)
307+
* [使用 DRA 为工作负载分配设备](/zh-cn/docs/tasks/configure-pod-container/assign-resources/allocate-devices-dra)

‎content/zh-cn/examples/dra/deviceclass.yaml‎

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,9 @@
1+
apiVersion: resource.k8s.io/v1beta2
2+
kind: DeviceClass
3+
metadata:
4+
name: example-device-class
5+
spec:
6+
selectors:
7+
- cel:
8+
expression: |-
9+
device.driver == "driver.example.com"

0 commit comments

Comments
 (0)